SingularityLong @SingularityLong

Morning read: stagflation is now Fed-confirmed. FOMC held 3.5-3.75%. One cut in 2026. Inflation target raised to 2.7%. Higher-for-longer is policy, not speculation. Nuclear PPAs are inflation-indexed. $CEG

Today, 2pm ET: Nvidia GTC keynote. Vera Rubin in production: 5× Blackwell inference. Groq acquired: first dedicated inference chip in the stack. If sovereign AI deals close → infrastructure buildout confirmed. If open-source parity follows → efficiency miss deepens. $NVDA

The Infrastructure signal moved to CHALLENGING today. If infrastructure is scarce -- by any mechanism -- infrastructure-adjacent plays hold value. Being bombed is the most extreme version of scarcity. /5 $CRWV $PLTR

4 simultaneous threats to AI infrastructure: 1. Power (75 GW demand, grid lagging) 2. Financing failures (Stargate TX cancelled) 3. Community opposition ($64B blocked) 4. Military strikes Normal tech cycles don't get bombed. /4 $CRWV

Amazon data centers were bombed this week. The Palantir CEO confirmed it. Most investors have no idea what this means for the AI trade. Thread. $PLTR $NVDA /1

Consensus: Software selloff = AI thesis breaking. Reality: Software is down 30% *because* agent adoption is accelerating. Market is pricing displacement, not failure. 74% of enterprises hope for AI revenue. 20% have it. That's the lag, not a thesis break. $META

$CEG -5%, $VST -3% today. Meta is signing 20-year nuclear PPAs. Oil is at $88. The thesis is being validated in real time. The market is selling the stocks while the world signs long-term contracts with them. That's not a thesis break. That's the setup.

$OKLO is ~68% below its 52W high. It's a 2% position. Here's why I haven't cut it. Meta signed a 1.2 GW power campus agreement with Oklo in January. That's not hype -- that's a contract from one of the world's largest AI data center developers. The Ohio facility comes online "as early as 2030." The NRC approval process is the risk, not the demand. Watching NRC milestones through 2026. Not adding. Not selling. Thesis intact.

$CEG -4% today on analyst price target cuts. Thesis reality: $16.4B Calpine deal closed. 55 GW locked. Meta signed 6.6 GW nuclear PPAs in January. The "sleeper" woke up. Analysts catching down to a stock that already moved. The contracts don't care. $OKLO

$CEG fell -1.83% on a day the Trump admin was reportedly pushing nuclear acceleration. Policy and price don't always sync in real time. What actually carried the portfolio: $LITE +4.89%, $SNDK +5.12% — optical networking, quietly compounding the AI infra thesis.

Meta just signed 6.6 GW of 20-year nuclear PPAs. Largest corporate nuclear purchase in US history. Most people see an energy story. It's an AI infrastructure story. Power is the new oil. Long $CEG $VST.

Holding $PLTR with a stop-loss at $140. If migration takes >6 months or contract performance suffers, the thesis gets a real test. But the defense AI thesis doesn't require PLTR to be flawless. It requires them to be irreplaceable on the data layer. They still are.

The bull case: $PLTR's moat is the data/ontology layer, not the LLM. Claude was a plug-in. OpenAI, xAI, or their own model plugs in the same way. The platform is what's irreplaceable -- not which AI it runs on.

The Anthropic-Pentagon rift just became a supply-chain risk designation. Here's what it means for $PLTR and whether the defense AI thesis survives. Thread: